CN111750962B - A high-precision estimation method of object weight based on filtering - Google Patents

Abstract

The invention discloses a filtering-based high-precision estimation method for object weight, which belongs to the field of information technology. The state transition equation is established according to the law that the input weight changes with the acceleration of the motor, the observation equation is established according to the data measured by the gravity sensor, the k-th weight prediction value is obtained through the k-1 weight estimated value, and the Kalman gain is used to convert the The weighted average of the k-th predicted weight value and the k-th sensor measurement value is used to obtain the k-th preliminary weight estimate. When k is greater than or equal to L and less than M, the k-th preliminary weight estimate is smoothed and the new weight is calculated. The k-th weight estimate value after processing is substituted into the subsequent operation, and the cycle is ended when k is greater than or equal to M; the error between the final weight estimate value and the actual value is not more than 0.3% by the above method; and then the motor is controlled according to the estimate value. Stop, so that the feeder can achieve accurate weight feeding operation.

Description

A high-precision estimation method for object weight based on filtering

technical field

The invention relates to a filtering-based high-precision estimation method for object weight, and belongs to the field of information technology.

Background technique

Feeder is a kind of raw material adding equipment. In some industrial processes with harsh conditions and strict requirements on the speed and amount of raw materials, such as the smelting process, the raw materials need to be fed into a high-temperature furnace. On the one hand, the temperature is relatively high. , people can not approach, on the other hand, the thickness of the feeding layer and the feeding speed must be uniform, and the work cannot be completed by manpower, so a special equipment feeder is required.

During the feeding process of the feeder, it is necessary to determine the feeding time and the feeding weight according to the specific feeding requirements. Usually, the weighing sensor is used to measure the weight. When the weight reaches the set value, the feeder motor is given a signal to decelerate, thereby gradually stopping. Blanking.

In the weighing process of the load cell, its own accuracy will be affected by internal noise and external environmental interference, and the material placed under it will also bring a certain impact error to the sensor, thus affecting the measurement accuracy of the load cell. In addition, after the motor stops rotating, there will still be a part of the remaining material in the air or at the feeding port will continue to be released. The errors caused by these factors will affect the measurement accuracy of the load cell from many aspects, that is to say, there is a certain error between the actual measurement value of the load cell at any time and the actual feeding value of the feeder at this time. If the measurement value of the load cell is used to control the motor of the feeder to decelerate or stop, it will inevitably cause the feeder to fail to perform accurate weight feeding.

Therefore, it is necessary to estimate the actual feeding value of the feeding machine at each moment in the feeding process, so that the estimated value is as close to the actual feeding value of the feeding machine at each moment as possible, and then control the motor of the feeding machine to decelerate or stop according to the estimated value, so that feeding The machine can carry out precise feeding.

SUMMARY OF THE INVENTION

In order to further improve the feeding weight accuracy of the feeding machine, the present invention provides a filtering-based high-precision estimation method for object weight, the method comprising:

S1 establishes the state transition equation and the observation equation according to the data in the feeding process of the existing feeder; wherein the data in the feeding process of the feeder includes the actual weight value of the feeder at each sampling time, the motor acceleration value at each sampling time, and the motor running value. The speed value and the weight value measured by the gravity sensor at each sampling time;

S2 obtains the weight prediction value at the k-th sampling time through the weight estimated value at the k-1 sampling time, wherein the weight estimated value at the first sampling time is the actual weight value that has been fed by the feeder at the initial moment;

S3 takes the weighted average of the k-th weight prediction value and the k-th sensor measurement value through the Kalman gain to obtain the k-th weight estimate value;

S4 performs smoothing processing on the weight estimated value of the Lth and later times according to the weight predicted value and estimated value obtained at the first L sampling times:

Compare the X∈{k,k-1,...,k-L+1}th weight prediction with the X∈{k,k-1,...,k-L+1}th weight estimate respectively The corresponding difference is recorded as ΔW=[ΔW(k),ΔW(k-1),...,ΔW(k-L+1)], remove a maximum value and a minimum value in ΔW to get a new ΔW ;

According to the closeness of the sampling time, the weighted average of ΔW after removing the maximum and minimum values is obtained to obtain the updated k-th weight estimate.

Optionally, the state transition equation and observation equation established in S1 are:

State transition equation:

W(k)=AW(k-1)+V(k)+e(k) (1)

The observation equation is as follows:

g(k)=CW(k)+h(k) (2)

Among them, W(k) represents the real weight of the total amount of material fed by the feeder in the kth sampling; A is the state transition matrix; W(k-1) represents the total amount of material fed by the feeder in the k-1th sampling The real weight of ; V(k) represents the speed of the motor running at the kth sampling:

V(k)=V ₀ +Bk (3)

e(k) represents the estimated noise, which represents the slight change in the weight value change process in the state transition equation;

g(k) represents the measurement weight measured by the sensor at the kth sampling; C is the measurement transition matrix; h(k) represents the observation noise, representing the real weight W(k) at the kth sampling and the kth sampling time The error of the measured weight g(k) measured by the sensor;

B is the acceleration matrix of the motor, and k is the sampling times.

Optionally, described S2 includes: Obtain the weight prediction value when sampling the kth time according to the following formula:

表示第k-1次重量估计值。Among them, W'(k) represents the kth weight prediction value,

Indicates the k-1th weight estimate.

Optionally, the S3 includes:

Obtain the error covariance matrix P'(k) between the predicted weight at the k-th sampling and the actual weight of the total amount fed by the feeder at the k-th sampling:

P'(k)=AP(k-1)A ^T +Q (5)

为估计噪声e(k)的协方差；Among them, P(k-1) represents the error covariance matrix between the k-1th estimated value and the real weight of the total amount of material fed by the feeder during the k-1th sampling;

is the covariance of the estimated noise e(k);

Obtain the k-th Kalman gain K'(k) according to P'(k) and P(k):

K'(k)=P'(k)C ^T [CP(k)C ^T +R] ^-1 (6)

Use this to obtain the k-th weight estimate

where R is the covariance of the observation noise h(k).

Optionally, the S3 obtains the kth weight estimate by taking the weighted average of the kth weight prediction value and the kth sensor measurement value through the Kalman gain, including:

When k is less than L, update the covariance P(k) and set k=k+1 and continue the calculation; if k is greater than or equal to L, first smooth the k-th weight estimate, and then update the covariance P(k) And let k=k+1 and continue the calculation. The update formula is as follows:

P(k)=(1-K'(k)C)P'(k-1) (8)

同样进行平滑处理；其中M为设定的采样次数的上限值。Optionally, in the method, for the kth weight estimated value obtained at each adoption time after the Lth sampling time and before the Mth sampling time

Smoothing is also performed; where M is the upper limit of the set sampling times.

Optionally, the S4 includes:

Set the proportional coefficient η=[η ₀ ,η ₁ ,...,η _L-3 ] that characterizes the weights,

Wherein, η ₀ ≥ η ₁ ... ≥ η _L-3 and η ₀ +η ₁ + ... + η _L-3 =1;

为：Then the new k-th weight estimate after smoothing

for:

Among them, ΔW ₁ ^T represents the transpose of the first row of the ΔW matrix, and ΔW ₂ ^T represents the transpose of the second row of the ΔW matrix.

量测转移矩阵

optional, state transition matrix

Measurement Transfer Matrix

取

观测噪声h(k)的协方差R取

Optionally, estimate the covariance of the noise e(k)

Pick

The covariance R of the observation noise h(k) is taken as

The present application also provides a method for controlling the total amount of material fed by a feeder. The method adopts the above-mentioned high-precision estimation method for object weight based on filtering to estimate the weight of the feeder, and then controls the total amount of material fed by the feeder according to the estimated weight of the feeder. .

The beneficial effects of the present invention are:

The invention discloses a filtering-based high-precision estimation method for object weight, which belongs to the field of information technology. The state transition equation is established according to the law that the input weight changes with the acceleration of the motor, the observation equation is established according to the data measured by the gravity sensor, the k-th weight prediction value is obtained through the k-1 weight estimation value, and the k-th weight prediction value is obtained through Kalman gain. The weighted average of the k-th predicted weight value and the k-th sensor measurement value is used to obtain the k-th preliminary weight estimate. When k is greater than or equal to L and less than M, the k-th preliminary weight estimate is smoothed and the new weight is calculated. The k-th weight estimate value after processing is substituted into the subsequent operation, and the cycle is ended when k is greater than or equal to M; the error between the final weight estimate value and the actual value is not more than 0.3% by the above method; and then the motor is controlled according to the estimate value. Stop, so that the feeder can achieve accurate weight feeding operation.

Description of drawings

In order to illustrate the technical solutions in the embodiments of the present invention more clearly, the following briefly introduces the accompanying drawings used in the description of the embodiments. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can also be obtained from these drawings without creative effort.

FIG. 1 is a flowchart of a filtering-based high-precision control method for weight measurement in an embodiment of the present invention.

FIG. 2 is an estimation diagram of the uniform acceleration motion of the feeder motor in an embodiment of the present invention.

FIG. 3 is an estimation diagram of uniform motion of a feeder motor in an embodiment of the present invention.

FIG. 4 is an estimation diagram of the uniform deceleration motion of the feeder motor in an embodiment of the present invention.

Detailed ways

In order to make the objectives, technical solutions and advantages of the present invention clearer, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.

The feeding machine is driven by the motor for feeding operation. During the feeding process, the running state of the motor usually goes through three stages of uniform acceleration, uniform speed and uniform deceleration. The technical solution of the present application provides a filter-based weight estimation method, which accurately estimates the feeding weight of the feeder at each stage, and then controls the motor to decelerate or stop according to the estimated actual feeding value, so that the feeder can achieve accurate weight. Feeding operation.

Example 1:

This embodiment provides a filtering-based high-precision estimation method for object weight, see FIG. 1 , the method includes:

Step 101: Establish a state transition equation according to the law that the weight input by the feeder changes with the acceleration of the motor; and establish an observation equation according to the data measured by the gravity sensor.

Specifically, the state transition equation and the observation equation are established according to the data in the feeding process of the existing feeder, wherein the data in the feeding process of the feeder include the actual weight value of the feeder at each sampling time, the motor acceleration value at each sampling time, the motor The running speed value and the weight value measured by the gravity sensor at each sampling time.

The established state transition equation is as follows:

W(k)=AW(k-1)+V(k)+e(k) (1)

The established observation equation is as follows:

g(k)=CW(k)+h(k) (2)

Among them, V(k) represents the speed of the motor running at the kth sampling, which is constructed as follows:

V(k)=V ₀ +Bk (3)

若电机已经在以速度v运行，接着要做匀速运动或匀减速运动，则

B为加速度矩阵，根据实际加速度值a设置其值为

V ₀ represents the initial speed of the motor operation, and its value is taken according to the operation requirements: for example, if the motor starts with uniform acceleration, then

If the motor is already running at the speed v, and then to make a uniform motion or uniform deceleration motion, then

B is the acceleration matrix, set its value according to the actual acceleration value a

C是量测转移矩阵，取

W(k) represents the real weight of the total amount of material fed by the feeder during the kth sampling; g(k) represents the measurement weight measured by the sensor during the kth sampling; k is the number of sampling times, k=1,2, 3...; A is the state transition matrix, take

C is the measurement transition matrix, take

取

h(k)表示观测噪声，表示第k次采样时真实重量W(k)与第k次采样时传感器测得的测量重量g(k)的误差；观测噪声的协方差R取

e(k) represents the estimated noise, represents the small change in the weight value change process in the state transition equation, and estimates the covariance of the noise

Pick

h(k) represents the observation noise, which represents the error between the real weight W(k) at the k-th sampling and the measured weight g(k) measured by the sensor at the k-th sampling; the covariance R of the observation noise is taken as

获取第k次重量预测值W'(k)。Step 102, pass the k-1th weight estimation

Obtain the k-th weight prediction value W'(k).

The formula is as follows:

Assuming that the first sampling is the moment when the feeder starts feeding, that is, the estimated weight of the first time is 0. According to formula (4), the predicted weight of the second time can be obtained, that is, the predicted weight of the second sampling time.

Step 103 , taking the weighted average of the k-th weight prediction value and the k-th sensor measurement value through the Kalman gain to obtain the k-th estimated weight value.

For example, after obtaining the weight prediction value at the second sampling time in step 102, the weight prediction value at the second sampling time and the sensor measurement value at the second sampling time are weighted and averaged to obtain the second weight estimate value;

Obtain the error covariance matrix P'(k) between the predicted weight at the k-th sampling and the real value (that is, the real weight of the total amount fed by the feeder at the k-th sampling):

P'(k)=AP(k-1)A ^T +Q (5)

Among them, P(k-1) represents the error covariance matrix between the k-1th estimated value and the real weight of the total amount of material fed by the feeder during the k-1th sampling;

Obtain the k-th Kalman gain K'(k) according to P'(k) and P(k):

K'(k)=P'(k)C ^T [CP(k)C ^T +R] ^-1 (6)

Use this to obtain the k-th weight estimate

Step 104, determine whether the value of k is greater than or equal to L.

L is a constant value obtained according to the demand, and the meaning in the calculation process is: take the number of L for smoothing. The larger the L value is, the closer the estimated data is to the real value, but it will also increase the amount of calculation and the possibility of overestimating.

If k is less than L, after updating the covariance P(k), set k=k+1 and continue the calculation. The update formula is as follows:

P(k)=(1-K'(k)C)P'(k-1) (8)

If k is greater than or equal to L, continue to the next step.

Step 105, determine whether the value of k is less than or equal to M.

进行平滑处理，If k is less than or equal to M, continue to estimate the kth weight

smoothing,

If k is greater than M, end the loop,

Among them, M is the upper limit of the sampling times, and the specific value can be determined according to the weight requirements of the feeding machine and the set total sampling times in practical applications.

Step 106: Compare the X∈{k,k-1,...,k-L+1}th weight prediction value with the X∈{k,k-1,...,k-L+1}th weight The estimated values correspond to the difference, denoted as ΔW=[ΔW(k),ΔW(k-1),...,ΔW(k-L+1)], remove a maximum value and a minimum value in ΔW to get new ΔW.

代入之后的运算。Step 107: Calculate the weighted average of ΔW after removing the maximum and minimum values according to the proximity of the sampling time, and update the k-th weight estimate

Operation after substitution.

The specific calculation is as follows:

Set the proportional coefficient η=[η ₀ ,η ₁ ,...,η _L-3 ] that characterizes the weight value, the weight value is the weight value, which represents the proportion of several data to be smoothed, and the data here is after the difference difference value. The weights can be set according to the empirical data or historical data of the actual system.

Wherein, η ₀ ≥ η ₁ ... ≥ η _L-3 and η ₀ +η ₁ + ... + η _L-3 =1;

为：Then the new k-th weight estimate after smoothing

for:

Among them, ΔW ₁ ^T represents the transpose of the first row of the ΔW matrix, and ΔW ₂ ^T represents the transpose of the second row of the ΔW matrix.

贴近于投料机第k次采样时所投料的真实值。After the above process, the kth estimated weight value obtained in this step

It is close to the real value of the material fed during the kth sampling of the feeder.

To sum up, the present application establishes a state transition equation according to the law that the input weight changes with the acceleration of the motor, establishes an observation equation according to the data measured by the gravity sensor, and obtains the kth weight prediction value through the k-1th weight estimation value. , through the Kalman gain, the weighted average of the k-th weight prediction value and the k-th sensor measurement value is obtained to obtain the k-th weight estimate value. When k is greater than or equal to 7 and less than M, the k-th weight estimate value is smoothed Processing, remove a maximum value and a minimum value in the X∈{k,k-1,...,k-6}th weight estimation value, calculate the weighted average according to the proximity of the sampling time, and apply the new processing The k-th estimated weight value is substituted into the subsequent operation. If k is greater than or equal to M, the cycle ends; the current problem of inaccurate weight measurement is solved; and then the motor is subsequently controlled to decelerate or stop according to the estimated actual feeding value, so that feeding The machine can realize accurate weight feeding operation.

Embodiment 2:

Taking the initial speed V ₀ as 10 r/s as an example, this embodiment provides a high-precision estimation method for weight measurement based on filtering, which includes estimating three stages of uniform acceleration, uniform speed and uniform deceleration of the motor. The feeding weights of the feeder in each stage of uniform speed and uniform deceleration are 745g, 490g and 235g respectively, and each stage is sampled 50 times; in practical application, the method can include:

In step 201, a state transition equation is established according to the law that the weight input by the feeder changes with the acceleration of the motor; an observation equation is established according to the data measured by the gravity sensor.

The state transition equation is as follows:

W(k)=AW(k-1)+V(k)+e(k) (1)

The observation equation is as follows:

g(k)=CW(k)+h(k) (2)

Among them, V(k) represents the speed of the motor running at the kth sampling, which is constructed as follows:

V(k)=V ₀ +Bk (3)

B为加速度矩阵，根据实际加速度值a设置其值为

本实施例对a的取值为0.2、0和-0.2。V ₀ represents the initial speed of motor operation, take

B is the acceleration matrix, set its value according to the actual acceleration value a

In this embodiment, the values of a are 0.2, 0, and -0.2.

C是量测转移矩阵，取

W(k) represents the real weight of the total amount of material fed by the feeder in the kth sampling, and g(k) represents the measured weight measured by the sensor in the kth sampling. k is the sampling times, k=1, 2, 3...50. A is the state transition matrix, take

C is the measurement transition matrix, take

取

h(k)表示观测噪声，观测噪声的协方差R取

e(k) represents the estimated noise, the covariance of the estimated noise

Pick

h(k) represents the observation noise, and the covariance R of the observation noise is taken as

获取第k次重量预测值W'(k)。Step 202, pass the k-1th weight estimation

Obtain the k-th weight prediction value W'(k).

The formula is as follows:

in,

Step 203 , taking the weighted average of the k-th weight prediction value and the k-th sensor measurement value through the Kalman gain to obtain the k-th estimated weight value.

Get the error covariance matrix P'(k) between the kth weight prediction and the true value:

P'(k)=AP(k-1)A ^T +Q (5)

Get the k-th Kalman gain K'(k):

K'(k)=P'(k)C ^T [CP(k)C ^T +R] ^-1 (6)

Use this to obtain the k-th weight estimate

Step 204, determine whether the value of k is greater than or equal to 7.

If k is less than 7, after updating the covariance P(k), set k=k+1 and continue the calculation. The update formula is as follows:

P(k)=(1-K'(k)C)P'(k-1) (8)

If k is greater than or equal to 7, continue to the next step.

Step 205 , determine whether the value of k is less than the upper limit value of 50 of the sampling times.

进行平滑处理，If k is less than or equal to 50, continue to estimate the kth weight

smoothing,

If k is greater than 50, end the loop,

Step 206: Corresponding the X∈{k,k-1,...,k-6}th weight prediction value and the X∈{k,k-1,...,k-6}th weight estimate value respectively Make a difference, denoted as ΔW=[ΔW(k),ΔW(k-1),...,ΔW(k-6)], remove a maximum value and a minimum value in ΔW to get a new ΔW.

代入之后的运算。Step 207: Calculate the weighted average of ΔW after removing the maximum and minimum values according to the proximity of the sampling time, and update the k-th weight estimate

Operation after substitution.

The specific calculation is as follows:

Set the proportionality coefficient η=[η ₀ ,η ₁ ,...,η ₄ ] to characterize the weight, take η=[0.3, 0.25, 0.2, 0.15, 0.1]

Satisfy η ₀ ≥ η ₁ ≥ η ₂ ≥ η ₃ ≥ η ₄ and η ₀ +η ₁ +η ₂ +η ₃ +η ₄ =1;

为：Then the new k-th weight estimate after smoothing

for:

Among them, ΔW ₁ ^T represents the transpose of the first row of the ΔW matrix, and ΔW ₂ ^T represents the transpose of the second row of the ΔW matrix.

Result analysis

When the motor is used for uniform acceleration motion with an acceleration of 0.2r/s ² , the estimated curve can be used as shown in Figure 2, and the data statistics are shown in Table 1:

Table 1: Statistics of uniform acceleration motion data

When the motor is moving at a constant speed with an acceleration of 0, the estimated curve can be used as shown in Figure 3, and the data statistics are shown in Table 2:

Table 2: Statistics of uniform motion data

When the motor is used for uniform acceleration motion with an acceleration of -0.2r/s2, the estimated curve can be used as shown in Figure 4, and the data statistics are shown in Table 3:

Table 3: Statistics of uniform deceleration motion data

Note: k starts from 1,

It can be seen that, according to the method of the present invention, the estimated value (that is, the filtering value in Table 1-3) and the real value (that is, the The errors of the true value) are -0.0128%, -0.0234% and 0.0930% respectively, and the estimated value of the total amount is not more than 0.3% different from the true value, achieving a high-precision estimation, that is, the estimation of the feeding weight of the feeding machine The error between the value and the actual feeding value is not more than 0.3%.

In addition, it can be seen from the three graphs that the filter estimation curve is relatively smooth, with minimal fluctuation and high stability.

The high-precision and high-stability estimated value provides a more accurate judgment for the issuance of the motor deceleration signal, thereby providing a guarantee for improving the accuracy of the feeding weight.

Some steps in the embodiments of the present invention may be implemented by software, and corresponding software programs may be stored in a readable storage medium, such as an optical disc or a hard disk.

The above are only preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included in the protection of the present invention. within the range.

Claims (8)

1. A filtering-based high-precision object weight estimation method is characterized by comprising the following steps:

s1: establishing a state transfer equation and an observation equation according to data in the feeding process of the existing feeder; the data in the feeding process of the batch feeder comprise actual weight values fed by the batch feeder at each sampling moment, motor acceleration values at each sampling moment, motor running speed values and weight values measured by the gravity sensors at each sampling moment;

s2: acquiring a weight predicted value in the k-th sampling through a weight estimated value in the k-1 th sampling, wherein the weight estimated value in the 1 st sampling is an actual weight value of fed materials of a feeder at an initial moment;

s3: weighting and averaging the kth weight predicted value and the kth sensor measured value through Kalman gain to obtain a kth weight estimated value;

s4: according to the weight predicted value and the estimated value obtained at the previous L times of sampling, smoothing the estimated values of the weights after the L times and the L times:

respectively making a difference between an Xth weight predicted value belonging to { k, k-1., k-L +1} and an Xth weight estimated value belonging to { k, k-1., k-L +1}, wherein the difference is marked as delta W [ [ delta W (k) ], delta W (k-1) ], the.

According to the proximity degree of the sampling time, calculating weighted average of the delta W after the maximum value and the minimum value are removed to obtain an updated kth weight estimation value;

the state transition equation and the observation equation established in S1 are respectively:

the state transition equation:

W(k)＝AW(k-1)+V(k)+e(k) (1)

the observation equation is as follows:

g(k)＝CW(k)+h(k) (2)

wherein W (k) represents the real weight of the total amount fed by the feeder at the k-th sampling; a is a state transition matrix; w (k-1) represents the real weight of the total amount fed by the feeder at the sampling time of k-1; v (k) represents the speed at which the motor runs at the kth sample:

V(k)＝V₀+Bk (3)

e (k) represents estimation noise and represents small disturbance in the process of weight value change in the state transition equation; v₀Representing an initial speed of operation of the motor;

g (k) represents the measured weight measured by the sensor at the time of the kth sampling; c is a measurement transfer matrix; h (k) represents observation noise, and represents the error between the true weight W (k) at the k-th sampling and the measured weight g (k) measured by the sensor at the k-th sampling;

b is an acceleration matrix of the motor, and k is the sampling frequency;

said S2 includes; obtaining a weight predicted value at the k sampling according to the following formula:

wherein W' (k) represents the predicted value of the k-th weight,

represents the weight estimate of the k-1 st time.

2. The method according to claim 1, wherein the S3 includes:

obtaining an error covariance matrix P' (k) between the weight prediction value at the k-th sampling time and the real weight of the total amount fed by the feeder at the k-th sampling time:

P′(k)＝AP(k-1)A^T+Q (5)

wherein P (k-1) represents an error covariance matrix between the k-1 th estimated value and the real weight of the total amount fed by the feeder at the k-1 th sampling; q is the covariance of the estimated noise e (k);

obtaining the K-th Kalman gain K '(K) according to P' (K) and P (K):

K′(k)＝P′(k)C^T[CP(k)C^T+R]^-1 (6)

thereby obtaining the k-th weight estimation value

Where R is the covariance of the observed noise h (k).

3. The method according to claim 2, wherein the step S3 is implemented by using a kalman gain to obtain the kth weight estimation value by taking a weighted average of the kth weight prediction value and the kth sensor measurement value, and the method comprises:

when k is smaller than L, updating the covariance P (k), then changing k to k +1 and continuing to calculate; if k is larger than or equal to L, smoothing the k-th weight estimation value, updating the covariance P (k), and calculating continuously after k is k + 1;

the update formula is as follows:

P(k)＝(1-K′(k)C)P′(k-1) (8)。

4. a method according to claim 3, wherein the k-th weight estimate is obtained for each sample time after the L-th sample time and before the M-th sample time in the method

Carrying out smoothing treatment in the same way; where M is the upper limit of the set number of samples.

5. The method according to claim 3, wherein the S4 includes:

setting a proportionality coefficient eta of the representation weight as [ eta [ ]₀，η₁，...，η_L-3]，

Wherein eta is₀≥η₁...≥η_L-3And η₀+η₁+...+η_L-3＝1；

The new k-th weight estimate after smoothing

Comprises the following steps:

wherein, Δ W₁ ^TRepresenting the transpose of the first row of the AW matrix, AWW₂ ^TRepresenting the transpose of the second row of the aw matrix.

6. The method of claim 5, wherein the state transition matrix

Measurement transfer matrix

7. Method according to claim 6, characterized in that the covariance Q of the estimated noise e (k) is taken

Covariance R of observed noise h (k)

8. A method for controlling the total amount of material fed by a batch feeder, wherein the method estimates the weight of material fed by the batch feeder by using the method for estimating the weight of an object based on filtering according to any one of claims 1 to 7, and controls the total amount of material fed by the batch feeder according to the estimated weight of material fed.

Images